POST /api/v1/store/social/channels/website
Authorization: Bearer <store JWT>
Permission: store.social_channels.manage

Body:
{
  "displayName": "Support Chat",
  "aiProviderCode": "anthropic",        // optional, default: anthropic
  "aiModelCode": "claude-sonnet-4-6",   // optional, default: claude-sonnet-4-6
  "systemPromptOverride": null          // optional
}

1. Verify store context (JWT → storeId)
2. EnsureCanConnectChannelAsync → check AllowsSocialMessaging + MaxSocialChannels
3. Check no existing Website channel for this store (409 if one already exists)
4. channelToken = Guid.NewGuid().ToString("N")
5. Create SocialChannel {
     ChannelType = Website,
     ExternalAccountId = channelToken,   ← this IS the token
     AccessTokenEncrypted = ""           ← no Meta token needed
   }
6. Return CreateWebsiteChannelResponse

{
  "channelId": "3fa85f64-...",
  "channelToken": "a1b2c3d4e5f6...",
  "embedSnippet": "<script src=\"https://api.example.com/widget.js\" data-channel=\"a1b2c3...\" async></script>"
}

POST /api/v1/widget/chat/{channelToken}/messages

Body:
{
  "sessionId": "visitor-uuid-or-fingerprint",
  "text": "Hello, I have a question about my order",
  "contactDisplayName": "Ahmed"          // optional, set once
}

1. Validate channelToken → load SocialChannel (ChannelType=Website, not deleted)
2. Check channel.IsEnabled (422 if disabled)
3. Upsert SocialConversation keyed on (SocialChannelId, sessionId):
     - First message: create conversation with Status=Open
     - Subsequent messages: update ContactDisplayName if not yet set
4. Create SocialMessage { Direction=Inbound, Status=Delivered }
5. If channel.IsEnabled && channel.IsAgentEnabled:
     Create AiMessageJob { Status=Pending } → AI will reply automatically
6. Return full conversation history (all messages for this sessionId)

{
  "conversationId": "guid...",
  "messages": [
    { "id": "...", "direction": "Inbound",  "contentText": "Hello...", "isAiGenerated": false, "createdAt": "..." },
    { "id": "...", "direction": "Outbound", "contentText": "Hi! How can I help?", "isAiGenerated": true, "createdAt": "..." }
  ]
}

GET /api/v1/widget/chat/{channelToken}/messages
  ?sessionId=visitor-uuid
  &after=2026-06-09T10:05:00Z    // optional — only return messages after this timestamp

1. Load channel by token
2. Load conversation by (channelId, sessionId)
3. If no conversation yet → return empty message list (not an error)
4. Query SocialMessages where ConversationId = conv.Id AND CreatedAt > after
5. Return messages ordered by CreatedAt ASC

AiMessageDispatchJob (runs every minute):

1. Load pending jobs for Website channels
2. Fetch last 20 messages as conversation history
3. Build system prompt (AiSystemPromptBuilder)
4. Call configured AI provider (Anthropic / OpenAI)
5. On success:
     - Create SocialMessage { Direction=Outbound, IsAiGenerated=true }
     - NO external API call (Website channel — DB write is the delivery)
     - Update AiMessageJob { Status=Sent }
     - SocialTokenTracker.RecordUsageAsync(inputTokens, outputTokens)
6. On failure:
     - Exponential backoff; max 3 attempts before Status=Dead

POST /api/v1/store/social/conversations/{conversationId}/messages
Authorization: Bearer <store JWT>
Permission: store.conversations.manage

Body: { "text": "Your order ships tomorrow!" }

1. Load conversation → verify it belongs to this store
2. Check channel.IsEnabled
3. ChannelType == Website → skip external API call
4. Create SocialMessage {
     Direction = Outbound,
     IsAiGenerated = false,
     SentByUserId = current.Id,
     Status = Sent
   }
5. Update conversation.LastMessageAt
6. Return SocialMessageResponse

Store admin                       Sendy API                        Customer widget
    │                                 │                                    │
    ├─ POST /channels/website ────────►│                                    │
    │◄──── { channelToken, snippet } ─┤                                    │
    │                                 │                                    │
    │  (pastes snippet into website)  │                                    │
    │                                 │                                    │
    │                                 │◄─── POST /widget/chat/{token}/messages ──┤
    │                                 │     { sessionId, text }            │
    │                                 │                                    │
    │                                 ├── upsert Conversation              │
    │                                 ├── save SocialMessage (Inbound)     │
    │                                 ├── queue AiMessageJob               │
    │                                 │──── return messages ──────────────►│
    │                                 │                                    │
    │                      [Hangfire - every minute]                       │
    │                                 │                                    │
    │                                 ├── call AI provider                 │
    │                                 ├── save SocialMessage (Outbound)    │
    │                                 │                                    │
    │                                 │◄── GET /widget/chat/{token}/messages?after=... ┤
    │                                 │──── return new outbound message ──►│
    │                                 │                                    │
    ├─ POST /conversations/{id}/messages ►│                                │
    │  { text: "Manual reply" }       │                                    │
    │                                 ├── save SocialMessage (Outbound)    │
    │◄──── SocialMessageResponse ─────┤                                    │
    │                                 │◄── GET /widget/chat/{token}/messages?after=... ┤
    │                                 │──── return manual reply ──────────►│

dotnet ef database update